Using neural networks with new morphological variables to recognize the number of jets in e+ e- reactions

In this work, we aim to construct a new set of variables, to recognize the number of jets produced in the e+ e- events. These so-called morphological variables usually used in image processing and recognition problems, are comparable to the classical sphericity, aplanarity etc.. The amelioration of the recognition efficiency is obtained thanks to the use of a back-propagation neural network. The survey first done on the generated Lund Monte Carlo events could be reinforced thereafter by taking into account the simulation of the ALEPH detector. The neural network performed on this later kind of events, successfully identifies the 4 classes of events (event with 2, 3, 4 jets or with an isotropic distribution (0 jets)). (author)